Korea's year-end tax-selling dip:
is it real, and does it bounce?
Every December the Korean financial press runs the same story: "Small caps slide as large holders sell to dodge capital-gains tax." And the same follow-up: "They come back after the cutoff." I counted all twelve year-ends from 2014 to 2025, every listed stock, to see whether either half is true.
01The rule, and why Korea has this quirk
Two words first if you do not trade Korea. KOSPI is the main board (Samsung Electronics, SK Hynix, the large caps); KOSDAQ is the secondary board, closer to the Nasdaq in spirit — smaller, more tech and biotech, more volatile. The regular session runs 09:00–15:30 Korea Standard Time (UTC+9).
Here is the quirk. In Korea, most individuals pay no capital-gains tax on listed stocks. The exception is a "large shareholder" — anyone whose holding in a single stock exceeds a threshold on the December 31 shareholder register. Settlement is T+2, so to be off that register you must sell by the second-to-last trading day of the year. I call that the cutoff day.
The threshold has moved around: it fell to 1 billion won (about US$700,000) per stock in 2020, then jumped to 5 billion won (about US$3.6 million) from the 2023 year-end. So a natural expectation is that the effect should have faded after 2023. I check that at the end.
- The dip: close 10 trading days before the cutoff → close on the cutoff day
- The bounce: buy at the open the day after the cutoff, sell at the close that day / open 1 day, 1 week, 2 weeks, 1 month later
- For comparison, buying at the open of the first trading day of January
- 0.22% deducted on every buy for commission and tax; nothing on the sell
Stocks are split into under 100 billion won (~US$70M) / 100bn–1tn / over 1 trillion won (~US$700M). The folk wisdom says the selling lands on small caps where wealthy individuals hold big stakes. The baseline throughout is the same stocks bought on any random day and held for the same period.
The median is the result exactly in the middle when every trade is lined up from worst to best — half did better, half did worse. Unlike the average, a few huge winners cannot drag it around.
02The dip: real, but small
| Market cap | Count | Share up | Average | Median |
|---|---|---|---|---|
| Under 100bn won | 8,188 | 40.9% | −0.81% | −0.84% |
| 100bn–1tn won | 13,494 | 47.0% | +0.50% | −0.11% |
| Over 1tn won | 2,933 | 53.0% | +0.85% | +0.44% |
| Any 10 days (baseline) | 6.1M | 45.3% | +0.45% | −0.36% |
The direction matches the story: small caps fall, large caps rise. Over the same ten days, stocks above 1 trillion won had a median of +0.44% and stocks below 100 billion had −0.84%.
But look at the size. The small-cap dip is half a point worse than any random ten days (−0.36%). In twelve years the small-cap median fell more than 3% over this window only twice: 2017 (−3.2%) and 2022 (−4.6%). It is not the collapse the headlines imply.
03The bounce: far bigger than the dip
| Hold | Win rate | Average | Median | |
|---|---|---|---|---|
| 1 week | Under 100bn won | 58.2% | +2.32% | +0.97% |
| 100bn–1tn | 56.2% | +2.49% | +0.82% | |
| Over 1tn | 47.3% | +0.80% | −0.22% | |
| Any random day | 44.4% | +0.02% | −0.41% | |
| 2 weeks | Under 100bn won | 64.0% | +4.01% | +2.10% |
| 100bn–1tn | 58.1% | +3.48% | +1.43% | |
| Over 1tn | 51.9% | +1.87% | +0.44% | |
| Any random day | 44.7% | +0.47% | −0.57% | |
| 1 month | Under 100bn won | 58.8% | +5.36% | +2.06% |
| 100bn–1tn | 52.5% | +3.77% | +0.70% | |
| Over 1tn | 50.4% | +1.77% | +0.20% | |
| Any random day | 44.7% | +0.67% | −0.88% |
Buy a small cap the morning after the cutoff and hold two weeks, and 64% of trades made money. Buy on a random day and it is 45%. A 19-point gap is larger than anything else I have found on this blog. Even after the 0.22% cost, the median is +2.1%.
The one-day version is nothing special: 45.5% for small caps against a 40.3% baseline. The effect builds over one to two weeks, not one day.
04A 0.5-point dip and a 2-point bounce do not match
If tax selling caused the dip, the bounce should be about the size of the dip. Instead the dip is half a point against the baseline and the bounce is over two points. Four times as much.
The answer is the calendar. The day after the cutoff is around December 28. Two weeks from there runs to mid-January, and small caps rising in early January is one of the oldest seasonal patterns in any market — the "January effect". It is sitting inside these numbers.
To separate the two, I measured buying at the open of the first trading day of January instead — skipping the last two days of December.
| Win rate | Average | Median | |
|---|---|---|---|
| Under 100bn won | 58.9% | +3.28% | +1.38% |
| 100bn–1tn | 50.4% | +1.71% | +0.06% |
| Over 1tn | 49.6% | +1.04% | −0.04% |
| Any random day | 44.7% | +0.47% | −0.57% |
Two days late, small caps still win 59% of the time. Entering right after the cutoff gave 64%. Those 5 points are roughly the tax-selling reversal; the other 14 points (59% minus the 45% baseline) are early-January small-cap strength.
The bounce the press describes is half tax, and more than half calendar.
05Year by year, and after the threshold went to 5 billion
| Year-end | Cutoff | Dip | 1-week bounce | Win rate |
|---|---|---|---|---|
| 2014 | Dec 26 | −0.42% | −0.22% | 45.4% |
| 2015 | Dec 28 | +1.40% | −0.22% | 45.1% |
| 2016 | Dec 27 | +0.99% | +2.46% | 72.0% |
| 2017 | Dec 26 | −3.21% | +2.41% | 69.5% |
| 2018 | Dec 26 | −1.80% | +0.75% | 57.0% |
| 2019 | Dec 26 | +1.31% | +0.41% | 53.1% |
| 2020 | Dec 28 | −2.15% | +4.87% | 82.8% |
| 2021 | Dec 28 | 0.00% | +1.00% | 60.6% |
| 2022 | Dec 27 | −4.56% | −2.78% | 22.5% |
| 2023 | Dec 26 | −1.04% | +1.19% | 64.5% |
| 2024 | Dec 26 | +0.17% | +3.33% | 79.6% |
| 2025 | Dec 26 | −1.58% | −0.14% | 47.9% |
Up in 8 of 12 years. 2014, 2015 and 2025 were flat, and 2022 was the one big loss: the deepest dip of the twelve (−4.6%) with no bounce at all, because KOSDAQ kept falling through the first week of January. When the tax selling lifts, the market still has to cooperate.
Then the years after the threshold rose to 5 billion won: 64.5%, 79.6%, 47.9%. It did not disappear. 2024 was the second-best year of the twelve. The pool of people who owed the tax shrank enormously, yet the bounce stayed — which points, as section 04 did, at January rather than tax as the main driver.
06Caveats
There are only twelve year-ends. Multiplying by stocks gives thousands of trades, but one year's market moves all of that year's trades together (2022 is the proof). "8 of 12" is the honest summary.
Market cap is approximate. I used today's share count times that year's cutoff-day close. Companies that issued or cancelled shares since may sit in the wrong bucket.
Delisted companies are not in the data. The small-cap bucket in particular may look better than reality.
Costs are the 0.22% buy-side deduction only. No sell-side cost and no bid-ask spread, and Korean small caps can have wide spreads.
07What I take from it
- Small caps do dip in the ten days before the cutoff — by about half a point more than any ten days
- Buying small caps the morning after the cutoff and holding two weeks won 64% of the time (baseline 45%)
- More than half of that bounce is the January effect: entering two days later in January still won 59%
- Up in 8 of 12 years; it failed when the market was falling anyway (2022)
- It did not fade after the tax threshold rose fivefold in 2023
Selling small caps in mid-December because "the tax selling is coming" looks, in the data, like giving up two points to save half a point. Buying the morning after the cutoff worked in 8 years out of 12 — but mostly because it is the start of January, not because of the tax, and January does not show up in a falling market.
The reason to sell in late December was weaker than advertised,
and the reason to buy in early January had little to do with tax.
댓글 쓰기